Generative AI is in use or planned at 90% of financial institutions surveyed in Hong Kong, up from 75% in 2025, according to a report released on November 17 by the Hong Kong Institute for Monetary and Financial Research (HKIMR, 香港貨幣及金融研究中心).
HKIMR is the research arm of the Hong Kong Academy of Finance, set up jointly by the HKMA, the SFC, the Insurance Authority and the Mandatory Provident Fund Schemes Authority. The report, titled From Assistance to Autonomy, draws on a survey and in-depth interviews with financial institutions and technology and regtech providers carried out between January and July 2026.
For agentic AI, meaning systems that carry out multi-step tasks with little human direction, 13% of institutions have already put it in place and 56% plan to. Respondents named explainability and security as the main obstacles. Enoch Fung (馮殷諾), chief executive of the academy, said at the report briefing, as quoted by Wen Wei Po, that explainability risk arises when staff cannot explain an AI output and so cannot take responsibility for it, and said training should cover accountability and communication as well as technical skills.
- Generative AI in use or planned: 90%, against 75% in 2025
- Agentic AI: 13% implemented, 56% planning
- Compliance: 86% use or plan generative AI tools
Why it matters: The academy is backed by all four Hong Kong financial regulators, so this survey is the closest thing to an official measure of AI use in the city’s financial sector. Compliance is where the move to agents raises the most questions: more than 70% of firms want to try agentic AI there, and an agent that screens or files without a person checking each step runs straight into the explainability problem Fung describes. Firms that deploy agents will have to show supervisors who answers for each output.
Sources: HKMA / HKIMR press release, November 17, 2026; 香港文匯報 / Wen Wei Po, November 18, 2026; Now財經台 / Now Finance, November 17, 2026 (Chinese-language sources, translated)
Image: Choinowski, CC BY-SA 4.0, via Wikimedia Commons




